Expression of random analytical error as a precentage of the range of clinical interest.

摘要

I propose that random method error be expressed as a percentage of the range(s) of clinical interest (upper range limit minus lower range limit), in addition to its usual forms of presentation. The interpretation of this percent random error is simple because a given percentage error implies the same reliability, regardless of test or range of clinical interest. Two different upper limits for percent random analytical error are suggested, based on the medical application of the test and the statistical properties of various amounts of error. When the purpose of the test is to distinguish an upper medically defined limit from a lower medically defined limit, a maximum random error equivalent to 60% of the difference between the limits is proposed. If the purpose of the test is to monitor values in the neighborhood of a normal range or therapeutic range, an upper limit for random error equal to 20% of the difference between the limits of the range is suggested.